
# http://proceedings.mlr.press/v101/huang19a/huang19a.pdf
# https://www.researchgate.net/publication/220875351_Generative_Models_for_Labeling_Multi-object_Configurations_in_Images
# https://www.tensorflow.org/datasets/catalog/open_images_v4
# Auto-Encoding Progressive Generative Adversarial Networks For 3D Multi Object Scenes


%config Completer.use_jedi = False
from ipywidgets import IntProgress
import matplotlib.pyplot as plt
from tensorflow.keras import layers, losses
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
import logging
import tensorflow_datasets as tfds
import pandas as pd
from tqdm import tqdm_notebook as tqdm
from sklearn.mixture import GaussianMixture
import os

seed = 1
np.random.seed(1)
tf.random.set_seed(1)


batch_size = 32
epochs = 10
dataset_name = 'wider_face'


if dataset_name == 'bdd100k':
    train_ds = tf.keras.preprocessing\
        .image_dataset_from_directory(directory='../data/bdd100k/images/10k/train1/',batch_size=batch_size)# train
    test_ds = tf.keras.preprocessing\
        .image_dataset_from_directory(directory='../data/bdd100k/images/10k/test1/',batch_size=batch_size) # test
    validation_ds = tf.keras.preprocessing\
        .image_dataset_from_directory(directory='../data/bdd100k/images/10k/val1/',batch_size=batch_size) # validation
elif dataset_name in ['flic','fashion_mnist','mnist','kitti']:
    train_ds,test_ds = tfds.load(name=dataset_name,split=['train', 'test']\
                                               ,as_supervised=False,download=True)
    validation_ds = test_ds
elif dataset_name in ['wider_face']:
    train_ds,test_ds,validation_ds = tfds.load(name=dataset_name,split=['train', 'test','validation']\
                                               ,as_supervised=False,download=True)
else:
    raise ValueError(f'Unhandled dataset {dataset_name}')


if dataset_name == 'bdd100k':
    dims = [x[0].get_shape().as_list() for x in train_ds]
    dims_df= pd.DataFrame.from_records(data=dims,columns=['batch','height','width','depth'])
else:
    dims = [x['image'].get_shape().as_list() for x in train_ds]
    dims_df= pd.DataFrame.from_records(data=dims,columns=['height','width','depth'])
dims_df.describe()


m = 20
height = int(min(dims_df['height'])/m)*m
width = int(min(dims_df['width'])/m)*m
# height = 2**(int(np.log2(min(dims_df['height']))))
# width = 2**(int(np.log2(min(dims_df['width']))))
depth = min(dims_df['depth'])
height,width = min(height,width),min(height,width)
height,width,depth

(160, 160, 3)


for t in train_ds.take(3):
    print(t.keys())

dict_keys(['faces', 'image', 'image/filename'])
dict_keys(['faces', 'image', 'image/filename'])
dict_keys(['faces', 'image', 'image/filename'])


if dataset_name == 'bdd100k':
    train_ds = train_ds.map(lambda x0,x1: x0/255.)
    test_ds = test_ds.map(lambda x0,x1: x0/255.)
    validation_ds = validation_ds.map(lambda x0,x1: x0/255.)
else:
    
    train_ds = train_ds.map(lambda x: tf.image.resize(images=tf.cast(x['image'],dtype=tf.float32)/255.,\
                                                      size=[height,width]))
    train_ds = train_ds.batch(batch_size,drop_remainder=True)
    ###
    test_ds = test_ds.map(lambda x: tf.image.resize(tf.cast(x['image'],dtype=tf.float32)/255.,\
                                                    size=[height,width]))
    test_ds = test_ds.batch(batch_size,drop_remainder=True)
    ###
    validation_ds = validation_ds.map(lambda x: tf.image.resize(tf.cast(x['image'],dtype=tf.float32)/255.\
                                                                ,size=[height,width]))
    validation_ds = validation_ds.batch(batch_size,drop_remainder=True)

###
train_ds_double_zipped = tf.data.Dataset.zip(datasets=(train_ds,train_ds))
test_ds_double_zipped = tf.data.Dataset.zip(datasets=(test_ds,test_ds))
validation_ds_double_zipped = tf.data.Dataset.zip(datasets=(validation_ds,validation_ds))


latent_dim = 4096
class CAE(tf.keras.Model):
    """Convolutional variational autoencoder."""

    def __init__(self, latent_dim):
        super(CAE, self).__init__()
        self.latent_dim = latent_dim
        self.logger = logging.getLogger('CAE')
        self.encoder = tf.keras.Sequential(name='encoder',layers=\
            [
                tf.keras.layers.InputLayer(input_shape=(height, width, depth)),
                tf.keras.layers.Conv2D(
                    filters=32, kernel_size=3, strides=(2, 2), activation='relu'),
                tf.keras.layers.Conv2D(
                    filters=64, kernel_size=3, strides=(2, 2), activation='relu'),
                tf.keras.layers.Flatten(),
                # No activation
                tf.keras.layers.Dense(latent_dim),
            ]
        )

        self.decoder = tf.keras.Sequential(name='decoder',layers=\
            [
                tf.keras.layers.InputLayer(input_shape=(latent_dim,)),
                tf.keras.layers.Dense(units=int(height/4) * int(width/4) * 32, activation=tf.nn.relu),
                tf.keras.layers.Reshape(target_shape=(int(height/4), int(width/4), 32)),
                tf.keras.layers.Conv2DTranspose(
                    filters=64, kernel_size=3, strides=2, padding='same',
                    activation='relu'),
                tf.keras.layers.Conv2DTranspose(
                    filters=32, kernel_size=3, strides=2, padding='same',
                    activation='relu'),
                # No activation
                tf.keras.layers.Conv2DTranspose(
                    filters=depth, kernel_size=3, strides=1, padding='same'),
            ]
        )

        self.encoder.summary()
        self.decoder.summary()
        
        
    def call(self, x):
        encoded = self.encoder(x)
        decoded = self.decoder(encoded)
        return decoded


cae = CAE(latent_dim)
cae.compile(optimizer='adam', loss=losses.MeanSquaredError())

Model: "encoder"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (None, 79, 79, 32)        896       
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 39, 39, 64)        18496     
_________________________________________________________________
flatten (Flatten)            (None, 97344)             0         
_________________________________________________________________
dense (Dense)                (None, 1024)              99681280  
=================================================================
Total params: 99,700,672
Trainable params: 99,700,672
Non-trainable params: 0
_________________________________________________________________
Model: "decoder"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_1 (Dense)              (None, 51200)             52480000  
_________________________________________________________________
reshape (Reshape)            (None, 40, 40, 32)        0         
_________________________________________________________________
conv2d_transpose (Conv2DTran (None, 80, 80, 64)        18496     
_________________________________________________________________
conv2d_transpose_1 (Conv2DTr (None, 160, 160, 32)      18464     
_________________________________________________________________
conv2d_transpose_2 (Conv2DTr (None, 160, 160, 3)       867       
=================================================================
Total params: 52,517,827
Trainable params: 52,517,827
Non-trainable params: 0
_________________________________________________________________


model_file_path = f'./models/cae_dataset_{dataset_name}_z_dim_{latent_dim}_data_dim_{height}x{width}x{depth}'
print(f'model path = {model_file_path}')

model path = ./models/cae_dataset_wider_face_z_dim_1024_data_dim_160x160x3


if os.path.exists(model_file_path):
    print('loading saved model')
    cae = tf.keras.models.load_model(filepath=model_file_path)
else:
    print('building model')
    # use checkpoints to save model fitting progress
    # https://www.tensorflow.org/api_docs/python/tf/keras/callbacks/ModelCheckpoint
    checkpoint_filepath = './checkpoint'
    model_checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(
        filepath=checkpoint_filepath,
        save_weights_only=True,
        monitor='val_loss',
        mode='max',
        save_best_only=True)
    # Model weights are saved at the end of every epoch, if it's the best seen
    # so far.
    cae.fit(x=train_ds_double_zipped,validation_data=test_ds_double_zipped,epochs=epochs,\
            callbacks=[model_checkpoint_callback],use_multiprocessing=True)
    # The model weights (that are considered the best) are loaded into the model.
    cae.load_weights(checkpoint_filepath)
    print('saving model')
    cae.save(filepath=model_file_path)

building model
Epoch 1/10
402/402 [==============================] - 524s 1s/step - loss: 0.0799 - val_loss: 0.0392
Epoch 2/10
402/402 [==============================] - 2946s 7s/step - loss: 0.0379 - val_loss: 0.0314
Epoch 3/10
402/402 [==============================] - 1353s 3s/step - loss: 0.0293 - val_loss: 0.0272
Epoch 4/10
402/402 [==============================] - 4426s 11s/step - loss: 0.0255 - val_loss: 0.0238
Epoch 5/10
402/402 [==============================] - 2632s 7s/step - loss: 0.0231 - val_loss: 0.0223
Epoch 6/10
402/402 [==============================] - 2468s 6s/step - loss: 0.0214 - val_loss: 0.0211
Epoch 7/10
402/402 [==============================] - 592s 1s/step - loss: 0.0199 - val_loss: 0.0203
Epoch 8/10
402/402 [==============================] - 575s 1s/step - loss: 0.0188 - val_loss: 0.0194
Epoch 9/10
402/402 [==============================] - 546s 1s/step - loss: 0.0180 - val_loss: 0.0186
Epoch 10/10
402/402 [==============================] - 1522s 4s/step - loss: 0.0170 - val_loss: 0.0187
saving model
INFO:tensorflow:Assets written to: ./models/cae_dataset_wider_face_z_dim_1024_data_dim_160x160x3/assets

INFO:tensorflow:Assets written to: ./models/cae_dataset_wider_face_z_dim_1024_data_dim_160x160x3/assets


# create valdation dataset tensor
for e in validation_ds.take(1):
    initial_state = tf.zeros(dtype=tf.float32,shape=e.shape)
validation_ds_tensor = validation_ds.\
        reduce(initial_state=initial_state,reduce_func=lambda x,y: tf.concat(values=[x,y],axis=0))
validation_ds_tensor = validation_ds_tensor[batch_size:] # drop dummy initial state


# calculate loss, can be compare over different dataset due to data scaling from 0 to 1
y_predicted = cae.predict(validation_ds)
cae_loss = cae.loss(y_pred=y_predicted,y_true=validation_ds_tensor).numpy()
print(f'CAE loss for dataset {dataset_name} = {np.round(cae_loss,4)}')

CAE loss for dataset wider_face = 0.03909999877214432


# plot decoded images

for batch in validation_ds.take(2):
        z = cae.encoder(batch).numpy()
        decoded_imgs = cae.decoder(z).numpy()
        
        for i in range(batch.shape[0]):
            fig, (ax1, ax2) = plt.subplots(1, 2)
            ax1.imshow(batch[i])
            ax2.imshow(decoded_imgs[i])

WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/ipykernel_launcher.py:8: RuntimeWarning: More than 20 figures have been opened. Figures created through the pyplot interface (`matplotlib.pyplot.figure`) are retained until explicitly closed and may consume too much memory. (To control this warning, see the rcParam `figure.max_open_warning`).
  
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).


# getting z tensor
z_tensor = None

inf_or_unknown_cardinality = ((test_ds.cardinality()==tf.data.INFINITE_CARDINALITY)\
    or (test_ds.cardinality() == tf.data.UNKNOWN_CARDINALITY)).numpy()
batches = test_ds.cardinality().numpy() if not inf_or_unknown_cardinality else 500

with tqdm(total=batches) as pbar:
    for batch in test_ds.take(batches):
        z = cae.encoder(batch).numpy()
        if z_tensor is None:
            z_tensor = tf.convert_to_tensor(z)
        else:
            z_tensor = tf.concat([z_tensor,tf.convert_to_tensor(z)],axis=0)
        pbar.update(1)
        #print(f'z shape {z.shape}')
#         decoded_imgs = cae.decoder(z).numpy()
#         #print(f'decoded images shape {decoded_imgs[0].shape}')
#         plt.imshow(batch[0])
#         plt.show()
#         plt.imshow(decoded_imgs[0])
#         plt.show()
z_tensor.shape

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/ipykernel_launcher.py:8: TqdmDeprecationWarning: This function will be removed in tqdm==5.0.0
Please use `tqdm.notebook.tqdm` instead of `tqdm.tqdm_notebook`

TensorShape([16096, 1024])


z_np= z_tensor.numpy()
n_z = z_np.shape[0]
n_z_train = int(0.8*n_z)
z_train = z_np[:n_z_train]
z_test = z_np[n_z_train:]


random_state = 1
reg_covar = 0.1
logps = []
k_values = [1,10,20,50,70,80,100,200]
cov_types = ['diag','cov']
for k in k_values:
    for cov_type in ['diag','full']:
        try:
            gm_fit = GaussianMixture(n_components=k,covariance_type=cov_type,random_state=random_state,\
                                     reg_covar=reg_covar).fit(z_train)
            logp_gm = gm_fit.score(X=z_test)
            print(f'For Gaussin Mixture with k = {k} and cov type {cov_type}, logp =  {logp_gm}  ')
            logps.append({'k':k,'cov_type':cov_type,'logp':logp_gm})
            print('##############  ')
        except Exception as e:
            print(f'Catched expection {e}  ')

For Gaussin Mixture with k = 1 and cov type diag, logp =  -411.81840966578113  
##############  
For Gaussin Mixture with k = 1 and cov type full, logp =  158.8647315390892  
##############  
For Gaussin Mixture with k = 10 and cov type diag, logp =  -303.1187019192367  
##############  
For Gaussin Mixture with k = 10 and cov type full, logp =  160.84260497329603  
##############  
For Gaussin Mixture with k = 20 and cov type diag, logp =  -281.46386500449336  
##############  
For Gaussin Mixture with k = 20 and cov type full, logp =  161.0652689730855  
##############  
For Gaussin Mixture with k = 50 and cov type diag, logp =  -253.4661870891009  
##############

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
<ipython-input-17-139827c4bcf3> in <module>
      8         try:
      9             gm_fit = GaussianMixture(n_components=k,covariance_type=cov_type,random_state=random_state,\
---> 10                                      reg_covar=reg_covar).fit(z_train)
     11             logp_gm = gm_fit.score(X=z_test)
     12             print(f'For Gaussin Mixture with k = {k} and cov type {cov_type}, logp =  {logp_gm}  ')

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/sklearn/mixture/_base.py in fit(self, X, y)
    191         self
    192         """
--> 193         self.fit_predict(X, y)
    194         return self
    195 

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/sklearn/mixture/_base.py in fit_predict(self, X, y)
    243                 prev_lower_bound = lower_bound
    244 
--> 245                 log_prob_norm, log_resp = self._e_step(X)
    246                 self._m_step(X, log_resp)
    247                 lower_bound = self._compute_lower_bound(

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/sklearn/mixture/_base.py in _e_step(self, X)
    296             the point of each sample in X.
    297         """
--> 298         log_prob_norm, log_resp = self._estimate_log_prob_resp(X)
    299         return np.mean(log_prob_norm), log_resp
    300 

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/sklearn/mixture/_base.py in _estimate_log_prob_resp(self, X)
    501             logarithm of the responsibilities
    502         """
--> 503         weighted_log_prob = self._estimate_weighted_log_prob(X)
    504         log_prob_norm = logsumexp(weighted_log_prob, axis=1)
    505         with np.errstate(under='ignore'):

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/sklearn/mixture/_base.py in _estimate_weighted_log_prob(self, X)
    454         weighted_log_prob : array, shape (n_samples, n_component)
    455         """
--> 456         return self._estimate_log_prob(X) + self._estimate_log_weights()
    457 
    458     @abstractmethod

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/sklearn/mixture/_gaussian_mixture.py in _estimate_log_prob(self, X)
    694     def _estimate_log_prob(self, X):
    695         return _estimate_log_gaussian_prob(
--> 696             X, self.means_, self.precisions_cholesky_, self.covariance_type)
    697 
    698     def _estimate_log_weights(self):

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/sklearn/mixture/_gaussian_mixture.py in _estimate_log_gaussian_prob(X, means, precisions_chol, covariance_type)
    409         log_prob = np.empty((n_samples, n_components))
    410         for k, (mu, prec_chol) in enumerate(zip(means, precisions_chol)):
--> 411             y = np.dot(X, prec_chol) - np.dot(mu, prec_chol)
    412             log_prob[:, k] = np.sum(np.square(y), axis=1)
    413 

<__array_function__ internals> in dot(*args, **kwargs)

KeyboardInterrupt:


logps_df = pd.DataFrame.from_records(data=logps)
logps_df.sort_values(by='logp',ascending=False).reset_index()

	height	width	depth
count	12880.000000	12880.0	12880.0
mean	888.309627	1024.0	3.0
std	350.513446	0.0	0.0
min	171.000000	1024.0	3.0
25%	682.000000	1024.0	3.0
50%	760.000000	1024.0	3.0
75%	1024.000000	1024.0	3.0
max	9108.000000	1024.0	3.0